Coresistance to quaternary ammonium compounds in extended-spectrum beta-lactamase-producing Escherichia coli

Background and Aim: Extended-spectrum β-lactamases (ESBL) in Escherichia coli constitutes one of the major threats to modern medicine, and the increasing pollution with quaternary ammonium compounds (QACs) has been suspected to contribute to the spread of ESBL-producing bacteria. The aim of the study was to investigate ESBLA and ESBLM-C-producing E. coli isolates for their coresistance to QACs and their phylogeny isolated from a Swedish University Hospital. Materials and Methods: Coresistance in E. coli with production of ESBL enzymes of the type blaCTX-M (n=23) was compared to E. coli producing AmpC type ESBL enzymes blaCMY and blaDHA (n=27). All isolates were tested for susceptibility to antibiotics and QACs, and high-quality whole-genome sequences were analyzed for resistance determinants. Results: The plasmid-borne small multidrug resistance (SMR) efflux pump sugE(p) was solely present in bla CMY-producing E. coli (n=9), within the same genetic environment blaCMY–blc–sugE(p). Other small multidrug efflux pumps were found without association for ESBL-types: emrE (n=5) and the truncated qacE∆1 (n=18). Conclusion: Coresistance of ESBL enzymes and SMR efflux pumps in E. coli was common and might indicate that other substances than antibiotics contribute to the spread and emergence of antibiotic resistance.


Introduction
Antimicrobial resistance is an urgent global threat to public health with a high disease burden to humans [1]. Since the 2000s, public awareness has increased regarding the worrisome rapid spread of antibiotic resistance, and many efforts have been taken to slow this development down. Actions have been taken to combat well-known factors that drive antibiotic resistance, such as limit the overuse of antibiotics and improve hygiene in medical care, animal husbandry, and the community [2].
However, despite all efforts, the emergence and distribution of extended-spectrum β-lactamases (ESBL) in Escherichia coli and other Enterobacteriaceae that cause resistance to the 3 rd -generation cephalosporins ESBL-producing E. coli continue, and therefore, other potential driving factors like the extensive use of biocides have been discussed. Concerns regarding the potential risk of biocidal substances have been raised early [3]. Especially, the intense use of substances that belong to quaternary ammonium compounds (QACs) has frequently been stressed as a potential risk, especially as the effects of sublethal concentrations on bacterial populations are rather unknown [4,5]. They are heavily utilized as preservatives and fabric softeners and thus released in significant amounts into the water cycle [6]. Anthropogenic contamination of soils and water environments with QACs can significantly contribute to enrichment of mobile genetic elements involving resistance determinants to antibiotics and QACs [7,8].
A majority of the clinically recognized ESBL enzymes have been mobilized from chromosomes of bacteria populating soils and other wet environments. Some ESBL enzymes became well-adapted in E. coli, such as bla CTX-M enzymes mobilized from Kluyvera spp., bla CMY enzymes from Citrobacter spp., or bla DHA from Morganella morganii, and they disseminate mainly through plasmids and certain bacterial clones [9,10]. Some of these plasmid-borne ESBL enzymes can be inhibited by beta-lactamase inhibitors, a fact that is used to classify ESBL enzymes in clinical contexts [11]. It distinguishes among others classical ESBL A (like bla CTX-M ) that are inhibitable by beta-lactamase inhibitors and those that cannot be inhibited which are called miscellaneous ESBL M-C (like bla CMY or bla DHA ); the latter are also known as plasmid-borne AmpC beta-lactamases.
Coresistance to bla CTX-M and biocides has been documented for QACs, in bacterial isolates involved in hospital outbreaks and environment [12,13]. However, this is less well studied for ESBL M-C . Against the background, that ESBL M-C -producing E. coli have become an emerging problem, it seems needful to study the coresistance potential of QACs to ESBL M-C [10,14].
Thus, the aim of the study was to investigate ESBL A and ESBL M-C -producing E. coli isolates for their coresistance to QACs and their phylogeny isolated from a Swedish University Hospital.

Ethical approval
After retrieval of relevant information from the referral, all isolates were anonymized; therefore, no ethical approval was necessary.

Bacterial isolates
This study comprises a total of 58 urinary tract isolates producing ESBL enzymes that were collected between 2011 and 2016 at Uppsala University Hospital. Urinary tract samples were cultured quantitatively on blood agar plates and cystine-lactoseelectrolyte-deficient agar plates (Oxoid, UK), and species identification was done using standard laboratory procedures and automated species identifications systems Maldi TOF (Bruker Daltonics, USA). The study followed the recommendations of the Nordic Committee on Antimicrobial Susceptibility Testing (www.nordicast.org) for diagnostics and classification of ESBL enzymes: All isolates with reduced susceptibility to cefpodoxime were further investigated using a synergy test that assesses the inhibition of cefotaxime, ceftazidime, or cefepime by clavulanic acid. Beta-lactamases that can be inhibited by clavulanic acid classified as classical ESBL (ESBL A ), and presumed enzymes that cannot be inhibited classified as suspicious plasmid-borne AmpC beta-lactamases (ESBL M-C ). Suspicious ESBL M-C were further verified by polymerase chain reaction (PCR) for the presence of plasmid-borne enzymes of type bla CMY, bla MOXM, bla DHAM, and bla ACCM (for details see below). All isolates were frozen as glycerol stock at −80°C.
For the purpose of this study, E. coli isolates originating from urinary tract samples producing ESBL A and ESBL M-C were randomly chosen resulting in 30 isolates producing ESBL A and 28 isolates producing ESBL M-C . After excluding isolates originating from the same individual, 54 isolates were included in the study; thereafter, all isolates were anonymized.

QACs
Susceptibility testing for biocides was performed by determining the minimal inhibition concentration (MIC) according to ISO 20776-1:2006 with the modification that ISO-Sensitest Broth (Oxoid, UK) was used. A microdilution assay with a final volume of 100 μL was used to determine the MIC for the following substances (concentration ranges in parenthesis): Benzalkonium chloride (BAC, 4-128 mg/L) and cetyltrimethylammonium bromide (CTAB; 4-128 mg/L) (all Sigma-Aldrich, USA). All isolates with elevated MIC values (BAC ≥64 mg/L and CTAB ≥64 mg/L) were retested in macrodilution format (1 mL). Stock solutions were prepared freshly and inoculated with bacteria within 2 h after a serial dilution in respective range. A final inoculate of 5×10 5 CFU/mL was prepared from an overnight culture in 1.5 mL Luria-Bertani broth (Sigma-Aldrich, USA) in room atmosphere at 35°C. The MIC assays were incubated for 18-20 h in room atmosphere at 35°C and the MIC values were read as the lowest concentration yielding no visible growth. E. coli ATCC 25922, Enterobacter Available at www.onehealthjournal.org/Vol.6/No.2/7.pdf cloacae CCUG 38138, and Klebsiella pneumoniae ATCC 700603 were used as control strains.

DNA preparation and whole-genome sequencing
One colony of each isolate was incubated in 2 mL LB broth (Sigma-Aldrich, USA) for 8 h at 37°C in room atmosphere. DNA preparation was done using a Wizard ® Genomic DNA Purification Kit (Promega, USA) according to the manufacturer's recommendations for Gram-negative bacteria with the exception that DNA was rehydrated with 10 mM Tris-HCl (pH 8.0). The quality and quantity of the extracted DNA was controlled by gel electrophoresis, spectrophotometry (Nanodrop, Thermo Fisher), and Quant-iT dsDNA BR assay and a Qubit instrument (Invitrogen). After standardizing the DNA extracts, the samples were transferred to Oxford Genome Center for library preparation and whole-genome sequencing. Fragmented DNA was end-repaired, A-tailed, adapter-ligated, and amplified using Nextera DNA library Prep (Illumina, USA). Sequencing was done on an Illumina HiSeq4000 platform, generating 150 bp paired-end reads.

Sequence analysis
Paired-end reads of the isolates from all the datasets were assembled using VelvetOptimiser software (v2.2.4) with kmer lengths from 21 to 99 using default optimization functions. Species confirmation, determination of phylotypes, and multi-locus sequence typing were performed according to the seven gene Achtman scheme using the pipeline implemented in Enterobase [16]. A neighbor-joining tree was constructed for rMLST allele nucleotide sequences of the study isolates [17,18]. Concatenated sequences for the rMLST scheme were retrieved through BIGsDB, aligned with MAFFT (v7.271, https://mafft.cbrc.jp/ alignment/software/) and the tree was calculated using PHYLIP (v3.695, http://evolution.genetics.washington.edu/phylip.html). Paralogous loci were excluded (BACT000060, BACT000065) resulting in 51 concatenated ribosomal loci for the rMLST. The dataset was then bootstrapped 500 times with phylip seqboot followed by calculations of distance matrices with phylip dnadist and neighbor-joining trees with phylip neighbor and a consensus tree using phylip consense. Illumina short reads were mapped to the antimicrobial resistance determinants database ARG-Annot (version 3) and a custom gene database for known biocide-related resistance genes using srst2 (v0.2.0, https://github.com/katholt/srst2; supplementary material Table-S1). The result was interpreted as positive when the minimum coverage was over 90%, maximum divergence under 10%, and a maximum number of mismatches per read of 10 (default settings). All findings from short-read mapping were confirmed using nucleotide BLAST on draft genomes, using a word size of 10, match/mismatch scores of 1/−2, the gap cost was linear, and the filter was set for low complexity regions; results were considered positive when the identity of a hit was over 90% (https://blast. ncbi.nlm.nih.gov/). Replicon sites of suspected plasmids were typed using PlasmidFinder (www.genomicepidemiology.org, December 2019) and insertion sequences were determined using ISFinder (https:// www-is.biotoul.fr, December 2019). All data have been submitted to the European Nucleotide Archives and are available under the project reference number PRJEB17631, Table-S2 in the supplemental material for individual accession numbers.

Statistical analysis
The strength of the association between phenotypic resistance and resistance determinants was calculated with R and to each group, ESBL A and ESBL M-C were determined with the odds ratio and 95% confidence intervals. Associations with p<0.05 and a lower confidence interval >1 were considered as statistically significant. Graphical illustrations of the results were produced using the package ggplot2 as implemented R (R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org/, version 3.4.4, 2018). The phylogenetic tree with metadata was illustrated using iTOL version 5 [19].

General comments on the datasets
Three isolates were excluded from further analysis; in one case because of production of solely bla SHV-12 , and could, therefore, not be assigned to the groups of interest; and in two further isolates, no ESBL determinants could be found. Thus, the resulting group sizes were n=23 for ESBL A

Sequence data
Species verification on sequence data using the Enterobase pipeline confirmed the purity of the whole-genome extracts and that all isolates belonged to the species E. coli. The average coverage of the high-quality short reads from all collections was 118 (SD of ±74). Draft genomes were obtained for all included isolates resulting in a median contig number (>10,000 bp size) of 55 (range 36-179), the median N50 value was 171,769 (range 48,359-322,359), the median total length of nucleotides assembled in the draft genome (>0 bp) was 5,179,586 bp (range 4,677,941-9,341,633 bp).

Antibiotic resistance determinants
Besides ESBL, other beta-lactamases were detected: bla OXA-1 that were only found in ESBL Aproducing isolates ( ESBL M-C, p<0.005); and TEM-1D beta-lactamases in 17 isolates, without a statistically significant distribution between the collections. The plasmid-borne quinolone resistance determinants qnrB were found in five isolates and only in ESBL M-C -producing isolates, however, the association was not statistically significant. In contrast, the quinolone resistance determinant aac(6')Ib-cr was only found in ESBL A isolates ( (Figure-2).

Biocidal resistance determinants
The plasmid-borne small multidrug efflux pump sugE(p) was only found in ESBL M-C isolates (9/27 [33%] ESBL M-C vs. 0/23 [0%] ESBL A , p=0.002). All isolates positive to sugE(p) produced concomitantly bla CMY-2 and bla CMY-42 , and the isolates were distributed within nearly all phylogenetic groups. In all sugE(p) positive isolates, the gene was found in the same genetic environment bla CMY -blc-sugE mostly with transposable insertion sequences. Four isolates had transposable elements that showed similarity to ISEcp1 (IS1380 family) and one had an ISSbo1 (IS91 family) that was assembled on a contig with an IncI1 replicon site. For two isolates, no insertion sequences were found, and in further two, it was an incomplete ISEcp1 ( Figures-1 and 3). The  multidrug efflux pump emrE was found in five isolates, and the truncated qacE∆1 was found in 18 isolates; no association was seen for ESBL-types.
The multidrug efflux pump determinants acrABR, emrAB, and acrEF (envCD), and the outer membrane channel tolC were present in all isolates.

Discussion
The present study investigated coresistance to QACs in ESBL A -and ESBL M-C -producing E. coli isolates. None of the isolates showed increased tolerance to BAC or CTAB and compared to other studies, resistance determinants to QACs were rare. Nonetheless, ESBL M-C was associated to the plasmid-borne small multidrug resistance (SMR) efflux pump sugE(p), while ESBL A was mainly associated with other antibiotic resistance determinants that confer resistance to macrolides (mphA), chloramphenicol (cat), aminoglycosides (aac-aad), quinolones aac(6')lb-cr, and beta-lactamases bla OXA-1 . The different ESBL types were also more frequently harbored by isolates belonging to different phylogroups: ESBL M-C was more often found in none-B2 phylogroups, especially in phylogroups A and D, while ESBL A was frequently found in B2 phylogroup.
Major resistance mechanism to QACs in Gramnegative bacilli is mediated through efflux pumps, where SMR) proteins are known to confer resistance to a variety of QACs. The sugE(p) has been described to mediate tolerance to a range of antiseptics and other toxic lipophilic compounds [20]. None of the isolates in the study harboring plasmid-borne sugE(p) genes showed increased MIC values to CTAB or BAC. Chung and Saier [21] showed that overexpression of chromosomally encoded sugE determinant in E. coli does only confer phenotypic tolerance to a narrow spectrum of QACs including CTAB, which, in addition, might be induced by mutants in sugE leading to hypersensitivity to QACs. It is thereby possible that isolates need selective pressure by QAC exposure to express phenotypic resistance. Still, it is uncertain how additional sugE genes, gained through horizontal gene transfer, and might give additional benefit to the host organism. Indeed, our findings might be in line with the report of Kermani et al. [22], who found that the primal function of SMR proteins is guanidinium export, and only a limited portion of these proteins mediates multidrug efflux. QACs are heavily used as disinfectants in animal food production, and phenotypic resistance has been measured in exposed isolates from these environments. In contrast, clinical isolates might not be exposed to significant levels of QACs, as these compounds are rather toxic to humans. However, QAC-resistant isolates have been found in hospital environments and have been linked to spreading with significant mortality [12].
It has been suggested that plasmid-borne sugE(p) genes have been mobilized from C. freundii, together with bla CMY and the outer membrane lipoprotein Blc, an event that has been hypothesized to happened at least 6 times [23]. However, only nine out of 22 bla CTX-M -producing E. coli isolates showed the genetic structure bla CMY -blc-sugE(p), which was accompanied by transposable elements. These genetic elements have been found in a variety of other Enterobacteriaceae, such as Klebsiella oxytoca, Salmonella spp., or Shigella spp. [24]. Curiously, the chromosomal environment of sugE(c) in E. coli compromises the same genetic structure with ampC-blc-sugE, however, no transposon-like element was found close by. Mobilization events are common in bacteria, and species that inhabited soils and wet environments have frequently been the source for antimicrobial resistance determinants causing huge problems in clinical situations [23]. Even though SMR proteins do not seem to transfer measurable tolerance to QACs tested here, they have been associated with increased mobilization and spread of antimicrobial resistance in polluted environments [8].
Once antibiotic resistance determinants have been acquired by human pathogens like E. coli, they can successfully spread by clones. E. coli belonging to ST131 have emerged during the 2000s as a pandemic, hypervirulent, and multiresistant clone [25]. Isolates that belong to E. coli ST131 do often produce bla CTX-M-15 , which has also been found in the present study: A majority of the bla CTX-M -producing isolates belonged to ST131. In contrast, for bla CTX-M -producing isolates from Uppsala University Hospital, no strong association to a certain clone was observed, the isolates were rather evenly distributed over all phylogenetic groups. So far, bla CMY enzymes spread mostly polyclonally within E. coli, and extensive dissemination was rather linked to mobile genetic elements such as IncI1 plasmids and insertions sequence ISEcp1 [26,27]. For the present collection, sugE(p) was mainly found in the context of a transposable ISEcp1 element and in one case with an ISSbo1 element that was assembled on an IncI1 plasmid. Chiu et al. [28] showed that blc and sugE(p) might have a regulatory function for bla CMY , leading to downregulation or upregulation, respectively.

Conclusion
Resistance determinants associated with SMR proteins that have been associated to QAC resistance were frequently found in ESBL-producing isolates, although no phenotypic tolerance could be detected. While the biological role of many proteins belonging to the SMR efflux pumps is not elusively clear, their wide spread might indicate other sources for selective pressure than antibiotics. grant from the Center for Research and Development Gävleborg, Uppsala University, Gävle, Sweden. The generation of sequence data was financially supported by a grant from Afa Insurance, Sweden (Grant number 150411) and Alf-de-Ruvo Memorial Foundation, Sweden. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.